Identifying Digital Attributes from Multiple Attribute Groups Utilizing a Deep Cognitive Attribution Neural Network

2. The non-transitory computer-readable medium of claim 1, wherein utilizing the cognitive attribution neural network to generate the tags corresponding to the plurality of attributes from the plurality of attribute groups for the object portrayed in the digital image comprises utilizing a base localization neural network of the cognitive attribution neural network to generate one or more localization feature vectors based on the digital image.

3. The non-transitory computer-readable medium of claim 2, wherein utilizing the cognitive attribution neural network to generate the tags corresponding to the plurality of attributes from the plurality of attribute groups for the object portrayed in the digital image comprises generating, based on the one or more localization feature vectors, a plurality of localization feature embeddings for the plurality of attribute groups.

4. The non-transitory computer-readable medium of claim 3, wherein utilizing the cognitive attribution neural network to generate the tags corresponding to the plurality of attributes from the plurality of attribute groups for the object portrayed in the digital image comprises determining, utilizing a plurality of attribute group classifiers of the cognitive attribution neural network, a set of attributes for the plurality of attribute groups based on the plurality of localization feature embeddings.

5. The non-transitory computer-readable medium of claim 4, wherein each attribute group classifier of the plurality of attribute group classifiers is configured to generate a subset of attributes corresponding to an attribute group of the plurality of attribute groups.

6. The non-transitory computer-readable medium of claim 5, wherein each attribute group classifier corresponds to an attribute group and is trained to generate a predicted attribute from a unique set of attributes corresponding to each attribute group.

7. The non-transitory computer-readable medium of claim 6, wherein the base localization neural network comprises a plurality of alternating dilated convolution layers and inception layers.

9. The non-transitory computer-readable medium of claim 4, wherein the operations further comprise generating the plurality of localization feature embeddings for the plurality of attribute groups by applying a global average pooling layer of the cognitive attribution neural network to the one or more localization feature vectors.

12. The computer-implemented method of claim 11, wherein the base localization neural network comprises a plurality of alternating dilated convolution layers and inception layers.

16. The system of claim 15, wherein the cognitive attribution neural network comprises a base localization neural network and a plurality of attribute group classifiers, wherein the base localization neural network comprises a plurality of alternating dilated convolution layers and inception layers.

17. The system of claim 16, wherein the one or more processor devices are further configured to cause the system to generate one or more localization feature vectors by analyzing the digital image via the plurality of alternating dilated convolution layers and inception layers of the base localization neural network of the cognitive attribution neural network.

18. The system of claim 17, wherein the one or more processor devices are further configured to cause the system to generate a plurality of attribute localization feature embeddings for a plurality of attribute groups based on the one or more localization feature vectors.

20. The system of claim 16, wherein the base localization neural network comprises a plurality of channels corresponding to a plurality of attributes for the object of the digital image and each channel of the plurality of channels comprises a plurality of alternating dilated convolution layers and inception layers.